A CPU bug with undocumented opcodes, possibly Intel (scroll to p. 155) | Hacker News Reader

A CPU bug with undocumented opcodes, possibly Intel (scroll to p. 155)

blackhat.com

63 pointsbaybal219 comments

Related talk (relevant part starts at 38:43) https://www.youtube.com/watch?v=KrksBdWcZgQ

Quote from the talk: I don't want to make it sound like the sky is falling. This was found on one very esoteric processor that is not used in widespread production. I think it's mostly interesting from an academic perspective that we have a tool that is able to find these kinds of things now.

Sorry for the typo, it is p. 155.

For quite a few days, people were suspecting that to be an Intel bug. The desktop shown in the video was a Dell Precision one, which means Core or a Xeon.

Another suspected cpu is said to be Intel X1000, if the demo was not running on the desktop. This was because Intel has issued a non-public errata just for major embedded partners before.

From myself, I will add that another suspect is Intel 80579 which is a still sold Pentium M/Core 1 SoC

While the title is a bit misleading, the presentation suggests enough to say that common CPU designs and architectures should be revised. Undocumented alone should always raise red flags. Imagine your average safe or door lock has undocumented keys or means to be opened without you noticing. Or your car having undocumented extras.

Page 154 says nothing about Intel. Kindly stop feeding the FUD.

155 probably

Here's the page for all who wondered.

https://image.prntscr.com/image/KDuFv8gRQYeLJG3pO8wwmQ.png

Within the context of the prior and subsequent pages, it appears to be talking about AMD.

It's the page after that. And, no, it doesn't seem to be AMD when looking at the side of the grey bar censoring the vendor.

Check the next page. Page numbers are not always reported faithfully on all PDF readers.

Besides, a 150+ page document of processor bugs and arcana is hardly FUD.

I second that. CPU bugs are the worst you can have. Means that are able to provide insight into instruction space of common CPUs are more valuable then your average particle collider.

It's page 155 and they've blanked out the vendor.

The bar is too short for Transmeta and too broad for AMD. It could well be Intel, judging from their other slide using Intel in the same location and font.

The length of the bar tells you nothing, as it simply covers the "(redacted)" text underneath.

The bar is just the right width for "(redacted)", I wouldn't read into it.

I assume OP is meaning p156 (title slide, author slide, +154). That said, is this not so new news?

https://twitter.com/xoreaxeaxeax/status/888702029744324608?l...

https://www.theregister.co.uk/2015/08/11/memory_hole_roots_i...

Presumably it hasn't been disclosed yet which systems are affected?

It looks like it says a Transmeta bug ... wtf is that? who cares?

One page later: (redacted) hardware bugs.

Sounds like something that would be hopefully fixable in microcode.

If it's really about Intel here, at least this seems like something that could be mitigated with microcode update.

Assuming that you mean page 155: What is your evidence that it is a bug on Intel Core?

This is misleading. The author of the talk has not revealed the vendor.

Reply on news.ycombinator.com